Find What Works: Schools

← All Philly stories

PhiladelphiaMethodologyeasy

SPREE agrees with our DB on 616/616 cells — Philly's data pipeline is structurally clean

Thesis

Philly's (iii-a) SPREE spot-check shows 100% agreement at ±1.5pp tolerance across 616 paired PSSA/Keystone cells. Mean delta is -0.002, max absolute delta is 0.05. This is structurally tighter than NYC's analogous Snapshot agreement (62-73% across metrics) because Philly's data pipeline keeps SPREE and the public source files numerically aligned end-to-end. The reconciliation gap NYC has between published and ingested doesn't exist here.

Supporting findings

  • 616 paired (school, year, metric, subgroup) cells; 616 within ±1.5pp.
  • Mean Δ = -0.002, max abs Δ = 0.05.
  • NYC comparison: 62-73% agreement on its Snapshot spot check, with deltas ranging -13 to +19 pp.
  • Implication: Philly users can trust the served numbers match the family-facing SPREE.
  • The structural difference is itself the story: NYC has cohort-attribution divergence between its source and its Snapshot; Philly doesn't.

Reporting directions

  • Write up the validation methodology as a public methodology page (already in /philly/methodology).
  • Reproduce for 2022-23 SPREE — does the 100% agreement hold across years?
  • Methodology comparison piece: NYC vs Philly on validation-gap origin.

Methodology & replication recipe — the queries + sources behind the findings above.

More on methodology